NSHM23 Draft, Gridded Only Hazard Maps

Table Of Contents

• PGA (g), 2% in 50 year
  • Mean hazard maps and comparisons, PGA (g), 2% in 50 year
  • Median hazard maps and comparisons, PGA (g), 2% in 50 year
  • Percentile comparison maps, PGA (g), 2% in 50 year
  • Bounds, spread, and COV, PGA (g), 2% in 50 year
  • PGA (g), 2% in 50 year Logic Tree Comparisons
    • Regional Seismicity Rate, PGA (g), 2% in 50 year
    • Seismicity Declustering Algorithm, PGA (g), 2% in 50 year
    • Seismicity Smoothing Kernel, PGA (g), 2% in 50 year
    • Off Fault Mmax, PGA (g), 2% in 50 year
• PGA (g), 10% in 50 year
  • Mean hazard maps and comparisons, PGA (g), 10% in 50 year
  • Median hazard maps and comparisons, PGA (g), 10% in 50 year
  • Percentile comparison maps, PGA (g), 10% in 50 year
  • Bounds, spread, and COV, PGA (g), 10% in 50 year
  • PGA (g), 10% in 50 year Logic Tree Comparisons
    • Regional Seismicity Rate, PGA (g), 10% in 50 year
    • Seismicity Declustering Algorithm, PGA (g), 10% in 50 year
    • Seismicity Smoothing Kernel, PGA (g), 10% in 50 year
    • Off Fault Mmax, PGA (g), 10% in 50 year
• 1.0s SA, 2% in 50 year
  • Mean hazard maps and comparisons, 1.0s SA, 2% in 50 year
  • Median hazard maps and comparisons, 1.0s SA, 2% in 50 year
  • Percentile comparison maps, 1.0s SA, 2% in 50 year
  • Bounds, spread, and COV, 1.0s SA, 2% in 50 year
  • 1.0s SA, 2% in 50 year Logic Tree Comparisons
    • Regional Seismicity Rate, 1.0s SA, 2% in 50 year
    • Seismicity Declustering Algorithm, 1.0s SA, 2% in 50 year
    • Seismicity Smoothing Kernel, 1.0s SA, 2% in 50 year
    • Off Fault Mmax, 1.0s SA, 2% in 50 year
• 1.0s SA, 10% in 50 year
  • Mean hazard maps and comparisons, 1.0s SA, 10% in 50 year
  • Median hazard maps and comparisons, 1.0s SA, 10% in 50 year
  • Percentile comparison maps, 1.0s SA, 10% in 50 year
  • Bounds, spread, and COV, 1.0s SA, 10% in 50 year
  • 1.0s SA, 10% in 50 year Logic Tree Comparisons
    • Regional Seismicity Rate, 1.0s SA, 10% in 50 year
    • Seismicity Declustering Algorithm, 1.0s SA, 10% in 50 year
    • Seismicity Smoothing Kernel, 1.0s SA, 10% in 50 year
    • Off Fault Mmax, 1.0s SA, 10% in 50 year

PGA (g), 2% in 50 year

(top)

Download Mean Hazard CSVs: pga_TWO_IN_50.csv pga_TWO_IN_50_comp.csv

Mean hazard maps and comparisons, PGA (g), 2% in 50 year

(top)

Primary (NSHM23 Draft, Gridded Only) Weighted-Average	Comparison (NSHM18, Gridded Only) Weighted-Average

The following plots compare mean hazard between NSHM23 Draft, Gridded Only and a comparison model, NSHM18, Gridded Only. The top row gives hazard ratios, expressed as % change, and the bottom row gives differences.

The left column compares the mean maps directly, with the comparison model as the divisor/subtrahend. Warmer colors indicate increased hazard in NSHM23 Draft, Gridded Only relative to NSHM18, Gridded Only.

The right column shows where and by how much the comparison mean model (NSHM18, Gridded Only) is outside the distribution of values across all branches of the primary model (NSHM23 Draft, Gridded Only). Here, places that are zeros (light gray) indicate that the comparison mean hazard map is fully contained within the range of values in NSHM23 Draft, Gridded Only, cool colors indicate areas where the primary model is always lower than the comparison mean model, and warm colors areas where the primary model is always greater. Note that the color scales are reversed here so that colors are consistent with the left column even though the comparison model is now the dividend/minuend.

Mean % Change	Comparison Mean % Change From Extremes
Mean Difference	Comparison Mean Difference From Extremes

Median hazard maps and comparisons, PGA (g), 2% in 50 year

(top)

Primary Weighted-Median	Comparison Weighted-Median

This section is the same as above, but using median hazard maps rather than mean.

Median % Change	Comparison Median % Change From Extremes
Median Difference	Comparison Median Difference From Extremes

Percentile comparison maps, PGA (g), 2% in 50 year

(top)

The maps below show where the comparison (NSHM18, Gridded Only) model mean (left column) and median (right column) map lies within the primary model (NSHM23 Draft, Gridded Only) distribution. Areas where the comparison mean or median map is outside the primary model distribution are shown here in black regardless of if they are above or below.

Comparison Mean Percentile	Comparison Median Percentile

Branched-average hazard can be dominated by outlier branches. The map below on the left shows the percentile at which the primary model's mean map lies within its own full hazard distribution. Areas far from the 50-th percentile here are likely outlier-dominated and may show up in percentile comparison maps, even if mean hazard differences are minimal. Keep this in mind when evaluating the maps above, and the influence of individual logic tree branches by this metric. The right map show the ratio of mean to median hazard.

Note: The mean map here is computed directly from mean hazard curves, but the median map is taken as the median value of hazard maps across all branches (rather than first calculating median curves at each location), which might bias this comparison.

Mean Map Percentile	Mean vs Median

Bounds, spread, and COV, PGA (g), 2% in 50 year

(top)

The maps below show the range of values across all logic tree branches, the ratio of the maximum to minimum value, and the coefficient of variation (std. dev. / mean). Note that the minimum and maximum maps are not a result for any single logic tree branch, but rather the smallest or largest value encountered at each location across all logic tree branches.

Minimum	Maximum
Log10 (Max/Min)	COV

PGA (g), 2% in 50 year Logic Tree Comparisons

(top)

This section shows how hazard changes across branch choices at each level of the logic tree. The summary figures below show mean hazard on the left, and then ratios & differences between the mean map considering subsets of the model holding each branch choice constant, and the overall mean map.

Combined Summary Maps

Regional Seismicity Rate, PGA (g), 2% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Regional Seismicity Rate branch level.

Download Choice Hazard CSV: pga_TWO_IN_50_SeisRate.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 13.04%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
PrefSeis
LowSeis
HighSeis

The table below gives % change maps between each option, head-to-head.

Choice	Vs PrefSeis	Vs LowSeis	Vs HighSeis
PrefSeis
LowSeis
HighSeis

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

PrefSeis	LowSeis	HighSeis

Seismicity Declustering Algorithm, PGA (g), 2% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Seismicity Declustering Algorithm branch level.

Download Choice Hazard CSV: pga_TWO_IN_50_SeisDecluster.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 18.14%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
GK
NN
Reas

The table below gives % change maps between each option, head-to-head.

Choice	Vs GK	Vs NN	Vs Reas
GK
NN
Reas

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

GK	NN	Reas

Seismicity Smoothing Kernel, PGA (g), 2% in 50 year

(top)

This section shows how mean hazard varies accross 2 choices at the Seismicity Smoothing Kernel branch level.

Download Choice Hazard CSV: pga_TWO_IN_50_SeisSmooth.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 45.87%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
Adaptive
Fixed

The table below gives % change maps between each option, head-to-head.

Choice	Vs Adaptive	Vs Fixed
Adaptive
Fixed

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

Adaptive	Fixed

Off Fault Mmax, PGA (g), 2% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Off Fault Mmax branch level.

Download Choice Hazard CSV: pga_TWO_IN_50_MmaxOff.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 4.98%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
MMax7.3
MMax7.6
MMax7.9

The table below gives % change maps between each option, head-to-head.

Choice	Vs MMax7.3	Vs MMax7.6	Vs MMax7.9
MMax7.3
MMax7.6
MMax7.9

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

MMax7.3	MMax7.6	MMax7.9

PGA (g), 10% in 50 year

(top)

Download Mean Hazard CSVs: pga_TEN_IN_50.csv pga_TEN_IN_50_comp.csv

Mean hazard maps and comparisons, PGA (g), 10% in 50 year

(top)

Primary (NSHM23 Draft, Gridded Only) Weighted-Average	Comparison (NSHM18, Gridded Only) Weighted-Average

The following plots compare mean hazard between NSHM23 Draft, Gridded Only and a comparison model, NSHM18, Gridded Only. The top row gives hazard ratios, expressed as % change, and the bottom row gives differences.

The left column compares the mean maps directly, with the comparison model as the divisor/subtrahend. Warmer colors indicate increased hazard in NSHM23 Draft, Gridded Only relative to NSHM18, Gridded Only.

The right column shows where and by how much the comparison mean model (NSHM18, Gridded Only) is outside the distribution of values across all branches of the primary model (NSHM23 Draft, Gridded Only). Here, places that are zeros (light gray) indicate that the comparison mean hazard map is fully contained within the range of values in NSHM23 Draft, Gridded Only, cool colors indicate areas where the primary model is always lower than the comparison mean model, and warm colors areas where the primary model is always greater. Note that the color scales are reversed here so that colors are consistent with the left column even though the comparison model is now the dividend/minuend.

Mean % Change	Comparison Mean % Change From Extremes
Mean Difference	Comparison Mean Difference From Extremes

Median hazard maps and comparisons, PGA (g), 10% in 50 year

(top)

Primary Weighted-Median	Comparison Weighted-Median

This section is the same as above, but using median hazard maps rather than mean.

Median % Change	Comparison Median % Change From Extremes
Median Difference	Comparison Median Difference From Extremes

Percentile comparison maps, PGA (g), 10% in 50 year

(top)

The maps below show where the comparison (NSHM18, Gridded Only) model mean (left column) and median (right column) map lies within the primary model (NSHM23 Draft, Gridded Only) distribution. Areas where the comparison mean or median map is outside the primary model distribution are shown here in black regardless of if they are above or below.

Comparison Mean Percentile	Comparison Median Percentile

Branched-average hazard can be dominated by outlier branches. The map below on the left shows the percentile at which the primary model's mean map lies within its own full hazard distribution. Areas far from the 50-th percentile here are likely outlier-dominated and may show up in percentile comparison maps, even if mean hazard differences are minimal. Keep this in mind when evaluating the maps above, and the influence of individual logic tree branches by this metric. The right map show the ratio of mean to median hazard.

Note: The mean map here is computed directly from mean hazard curves, but the median map is taken as the median value of hazard maps across all branches (rather than first calculating median curves at each location), which might bias this comparison.

Mean Map Percentile	Mean vs Median

Bounds, spread, and COV, PGA (g), 10% in 50 year

(top)

The maps below show the range of values across all logic tree branches, the ratio of the maximum to minimum value, and the coefficient of variation (std. dev. / mean). Note that the minimum and maximum maps are not a result for any single logic tree branch, but rather the smallest or largest value encountered at each location across all logic tree branches.

Minimum	Maximum
Log10 (Max/Min)	COV

PGA (g), 10% in 50 year Logic Tree Comparisons

(top)

This section shows how hazard changes across branch choices at each level of the logic tree. The summary figures below show mean hazard on the left, and then ratios & differences between the mean map considering subsets of the model holding each branch choice constant, and the overall mean map.

Combined Summary Maps

Regional Seismicity Rate, PGA (g), 10% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Regional Seismicity Rate branch level.

Download Choice Hazard CSV: pga_TEN_IN_50_SeisRate.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 15.38%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
PrefSeis
LowSeis
HighSeis

The table below gives % change maps between each option, head-to-head.

Choice	Vs PrefSeis	Vs LowSeis	Vs HighSeis
PrefSeis
LowSeis
HighSeis

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

PrefSeis	LowSeis	HighSeis

Seismicity Declustering Algorithm, PGA (g), 10% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Seismicity Declustering Algorithm branch level.

Download Choice Hazard CSV: pga_TEN_IN_50_SeisDecluster.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 23.58%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
GK
NN
Reas

The table below gives % change maps between each option, head-to-head.

Choice	Vs GK	Vs NN	Vs Reas
GK
NN
Reas

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

GK	NN	Reas

Seismicity Smoothing Kernel, PGA (g), 10% in 50 year

(top)

This section shows how mean hazard varies accross 2 choices at the Seismicity Smoothing Kernel branch level.

Download Choice Hazard CSV: pga_TEN_IN_50_SeisSmooth.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 38.73%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
Adaptive
Fixed

The table below gives % change maps between each option, head-to-head.

Choice	Vs Adaptive	Vs Fixed
Adaptive
Fixed

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

Adaptive	Fixed

Off Fault Mmax, PGA (g), 10% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Off Fault Mmax branch level.

Download Choice Hazard CSV: pga_TEN_IN_50_MmaxOff.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 4.18%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
MMax7.3
MMax7.6
MMax7.9

The table below gives % change maps between each option, head-to-head.

Choice	Vs MMax7.3	Vs MMax7.6	Vs MMax7.9
MMax7.3
MMax7.6
MMax7.9

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

MMax7.3	MMax7.6	MMax7.9

1.0s SA, 2% in 50 year

(top)

Download Mean Hazard CSVs: 1.0s_TWO_IN_50.csv 1.0s_TWO_IN_50_comp.csv

Mean hazard maps and comparisons, 1.0s SA, 2% in 50 year

(top)

Primary (NSHM23 Draft, Gridded Only) Weighted-Average	Comparison (NSHM18, Gridded Only) Weighted-Average

The following plots compare mean hazard between NSHM23 Draft, Gridded Only and a comparison model, NSHM18, Gridded Only. The top row gives hazard ratios, expressed as % change, and the bottom row gives differences.

The left column compares the mean maps directly, with the comparison model as the divisor/subtrahend. Warmer colors indicate increased hazard in NSHM23 Draft, Gridded Only relative to NSHM18, Gridded Only.

The right column shows where and by how much the comparison mean model (NSHM18, Gridded Only) is outside the distribution of values across all branches of the primary model (NSHM23 Draft, Gridded Only). Here, places that are zeros (light gray) indicate that the comparison mean hazard map is fully contained within the range of values in NSHM23 Draft, Gridded Only, cool colors indicate areas where the primary model is always lower than the comparison mean model, and warm colors areas where the primary model is always greater. Note that the color scales are reversed here so that colors are consistent with the left column even though the comparison model is now the dividend/minuend.

Mean % Change	Comparison Mean % Change From Extremes
Mean Difference	Comparison Mean Difference From Extremes

Median hazard maps and comparisons, 1.0s SA, 2% in 50 year

(top)

Primary Weighted-Median	Comparison Weighted-Median

This section is the same as above, but using median hazard maps rather than mean.

Median % Change	Comparison Median % Change From Extremes
Median Difference	Comparison Median Difference From Extremes

Percentile comparison maps, 1.0s SA, 2% in 50 year

(top)

The maps below show where the comparison (NSHM18, Gridded Only) model mean (left column) and median (right column) map lies within the primary model (NSHM23 Draft, Gridded Only) distribution. Areas where the comparison mean or median map is outside the primary model distribution are shown here in black regardless of if they are above or below.

Comparison Mean Percentile	Comparison Median Percentile

Branched-average hazard can be dominated by outlier branches. The map below on the left shows the percentile at which the primary model's mean map lies within its own full hazard distribution. Areas far from the 50-th percentile here are likely outlier-dominated and may show up in percentile comparison maps, even if mean hazard differences are minimal. Keep this in mind when evaluating the maps above, and the influence of individual logic tree branches by this metric. The right map show the ratio of mean to median hazard.

Note: The mean map here is computed directly from mean hazard curves, but the median map is taken as the median value of hazard maps across all branches (rather than first calculating median curves at each location), which might bias this comparison.

Mean Map Percentile	Mean vs Median

Bounds, spread, and COV, 1.0s SA, 2% in 50 year

(top)

The maps below show the range of values across all logic tree branches, the ratio of the maximum to minimum value, and the coefficient of variation (std. dev. / mean). Note that the minimum and maximum maps are not a result for any single logic tree branch, but rather the smallest or largest value encountered at each location across all logic tree branches.

Minimum	Maximum
Log10 (Max/Min)	COV

1.0s SA, 2% in 50 year Logic Tree Comparisons

(top)

This section shows how hazard changes across branch choices at each level of the logic tree. The summary figures below show mean hazard on the left, and then ratios & differences between the mean map considering subsets of the model holding each branch choice constant, and the overall mean map.

Combined Summary Maps

Regional Seismicity Rate, 1.0s SA, 2% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Regional Seismicity Rate branch level.

Download Choice Hazard CSV: 1.0s_TWO_IN_50_SeisRate.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 22.46%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
PrefSeis
LowSeis
HighSeis

The table below gives % change maps between each option, head-to-head.

Choice	Vs PrefSeis	Vs LowSeis	Vs HighSeis
PrefSeis
LowSeis
HighSeis

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

PrefSeis	LowSeis	HighSeis

Seismicity Declustering Algorithm, 1.0s SA, 2% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Seismicity Declustering Algorithm branch level.

Download Choice Hazard CSV: 1.0s_TWO_IN_50_SeisDecluster.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 15.42%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
GK
NN
Reas

The table below gives % change maps between each option, head-to-head.

Choice	Vs GK	Vs NN	Vs Reas
GK
NN
Reas

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

GK	NN	Reas

Seismicity Smoothing Kernel, 1.0s SA, 2% in 50 year

(top)

This section shows how mean hazard varies accross 2 choices at the Seismicity Smoothing Kernel branch level.

Download Choice Hazard CSV: 1.0s_TWO_IN_50_SeisSmooth.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 28.11%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
Adaptive
Fixed

The table below gives % change maps between each option, head-to-head.

Choice	Vs Adaptive	Vs Fixed
Adaptive
Fixed

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

Adaptive	Fixed

Off Fault Mmax, 1.0s SA, 2% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Off Fault Mmax branch level.

Download Choice Hazard CSV: 1.0s_TWO_IN_50_MmaxOff.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 14.78%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
MMax7.3
MMax7.6
MMax7.9

The table below gives % change maps between each option, head-to-head.

Choice	Vs MMax7.3	Vs MMax7.6	Vs MMax7.9
MMax7.3
MMax7.6
MMax7.9

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

MMax7.3	MMax7.6	MMax7.9

1.0s SA, 10% in 50 year

(top)

Download Mean Hazard CSVs: 1.0s_TEN_IN_50.csv 1.0s_TEN_IN_50_comp.csv

Mean hazard maps and comparisons, 1.0s SA, 10% in 50 year

(top)

Primary (NSHM23 Draft, Gridded Only) Weighted-Average	Comparison (NSHM18, Gridded Only) Weighted-Average

The following plots compare mean hazard between NSHM23 Draft, Gridded Only and a comparison model, NSHM18, Gridded Only. The top row gives hazard ratios, expressed as % change, and the bottom row gives differences.

The left column compares the mean maps directly, with the comparison model as the divisor/subtrahend. Warmer colors indicate increased hazard in NSHM23 Draft, Gridded Only relative to NSHM18, Gridded Only.

The right column shows where and by how much the comparison mean model (NSHM18, Gridded Only) is outside the distribution of values across all branches of the primary model (NSHM23 Draft, Gridded Only). Here, places that are zeros (light gray) indicate that the comparison mean hazard map is fully contained within the range of values in NSHM23 Draft, Gridded Only, cool colors indicate areas where the primary model is always lower than the comparison mean model, and warm colors areas where the primary model is always greater. Note that the color scales are reversed here so that colors are consistent with the left column even though the comparison model is now the dividend/minuend.

Mean % Change	Comparison Mean % Change From Extremes
Mean Difference	Comparison Mean Difference From Extremes

Median hazard maps and comparisons, 1.0s SA, 10% in 50 year

(top)

Primary Weighted-Median	Comparison Weighted-Median

This section is the same as above, but using median hazard maps rather than mean.

Median % Change	Comparison Median % Change From Extremes
Median Difference	Comparison Median Difference From Extremes

Percentile comparison maps, 1.0s SA, 10% in 50 year

(top)

The maps below show where the comparison (NSHM18, Gridded Only) model mean (left column) and median (right column) map lies within the primary model (NSHM23 Draft, Gridded Only) distribution. Areas where the comparison mean or median map is outside the primary model distribution are shown here in black regardless of if they are above or below.

Comparison Mean Percentile	Comparison Median Percentile

Branched-average hazard can be dominated by outlier branches. The map below on the left shows the percentile at which the primary model's mean map lies within its own full hazard distribution. Areas far from the 50-th percentile here are likely outlier-dominated and may show up in percentile comparison maps, even if mean hazard differences are minimal. Keep this in mind when evaluating the maps above, and the influence of individual logic tree branches by this metric. The right map show the ratio of mean to median hazard.

Note: The mean map here is computed directly from mean hazard curves, but the median map is taken as the median value of hazard maps across all branches (rather than first calculating median curves at each location), which might bias this comparison.

Mean Map Percentile	Mean vs Median

Bounds, spread, and COV, 1.0s SA, 10% in 50 year

(top)

The maps below show the range of values across all logic tree branches, the ratio of the maximum to minimum value, and the coefficient of variation (std. dev. / mean). Note that the minimum and maximum maps are not a result for any single logic tree branch, but rather the smallest or largest value encountered at each location across all logic tree branches.

Minimum	Maximum
Log10 (Max/Min)	COV

1.0s SA, 10% in 50 year Logic Tree Comparisons

(top)

This section shows how hazard changes across branch choices at each level of the logic tree. The summary figures below show mean hazard on the left, and then ratios & differences between the mean map considering subsets of the model holding each branch choice constant, and the overall mean map.

Combined Summary Maps

Regional Seismicity Rate, 1.0s SA, 10% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Regional Seismicity Rate branch level.

Download Choice Hazard CSV: 1.0s_TEN_IN_50_SeisRate.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 23.30%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
PrefSeis
LowSeis
HighSeis

The table below gives % change maps between each option, head-to-head.

Choice	Vs PrefSeis	Vs LowSeis	Vs HighSeis
PrefSeis
LowSeis
HighSeis

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

PrefSeis	LowSeis	HighSeis

Seismicity Declustering Algorithm, 1.0s SA, 10% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Seismicity Declustering Algorithm branch level.

Download Choice Hazard CSV: 1.0s_TEN_IN_50_SeisDecluster.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 21.21%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
GK
NN
Reas

The table below gives % change maps between each option, head-to-head.

Choice	Vs GK	Vs NN	Vs Reas
GK
NN
Reas

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

GK	NN	Reas

Seismicity Smoothing Kernel, 1.0s SA, 10% in 50 year

(top)

This section shows how mean hazard varies accross 2 choices at the Seismicity Smoothing Kernel branch level.

Download Choice Hazard CSV: 1.0s_TEN_IN_50_SeisSmooth.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 27.37%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
Adaptive
Fixed

The table below gives % change maps between each option, head-to-head.

Choice	Vs Adaptive	Vs Fixed
Adaptive
Fixed

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

Adaptive	Fixed

Off Fault Mmax, 1.0s SA, 10% in 50 year

(top)

This section shows how mean hazard varies accross 3 choices at the Off Fault Mmax branch level.

Download Choice Hazard CSV: 1.0s_TEN_IN_50_MmaxOff.csv

The table below gives summary statistics for the spatial average difference and average absolute difference of hazard between mean hazard maps for each individual branch choices. In other words, it gives the expected difference (or absolute difference) between two models if you picked a location at random. Values are listed between each pair of branch choices, and also between that choice and the overall mean map in the first column.

The overall average absolute difference between the map for any choice to each other choice, a decent summary measure of how much hazard varies due to this branch choice, is: 9.04%

The map table below shows how the mean map for each branch choice compares to the overall mean map, expressed as % change (first column) and difference (second column). The third column, 'Choice Percentile in Full Dist', shows at what percentile the map for that branch choice lies within the full distribution, and the fourth column, 'Choice Percentile in Dist Without', shows the same but for the distribution of all other branches (without this choice included).

Note that these percentile comparisons can be outlier dominated, in which case even if a choice is near the overall mean hazard it may still lie far from the 50th percentile (see 'Mean Map Percentile' above to better understand outlier dominated regions).

	Choice Mean vs Full Mean, % Change	Choice Mean - Full Mean	Choice Percentile in Full Dist	Choice Percentile in Dist Without
MMax7.3
MMax7.6
MMax7.9

The table below gives % change maps between each option, head-to-head.

Choice	Vs MMax7.3	Vs MMax7.6	Vs MMax7.9
MMax7.3
MMax7.6
MMax7.9

The table below shows how much the mean hazard map would change if each branch were eliminated. This differs from the above comparisons in that it also reflects the weight assigned to each branch. The sign is now flipped such that blue and green areas indicate areas where hazard is higher due to inclusion of the listed listed choice, and would go down were that choice eliminated.

MMax7.3	MMax7.6	MMax7.9